Meta pauses controversial employee surveillance program after data lea

Meta has indefinitely paused its controversial Model Capability Initiative (MCI) program after a security breach exposed sensitive employee laptop tracking databases to anyone inside the company. The program was designed to scrape US employee screens, mouse clicks, and keystrokes to train AI systems to use software like humans.

According to internal documents, the leak exposed employee data across 45,000 hive tables, including private chats, full prompts, transcriptions, and performance reviews. Meta CTO Andrew Bosworth admitted the breach was caused by misconfigured access control lists (ACLs) and promised a full investigation into how the privacy safeguards failed.

The tracking program had previously triggered widespread employee protests, with more than 1,600 workers signing an internal petition warning of security risks. Meta CEO Mark Zuckerberg had defended the program, arguing that AI models need to learn from the company's "smart people" rather than external contractors. Meta VP Stephane Kasriel confirmed the program would remain paused until data controls are verified.